Create metric(s) to estimate verbatim copy

Abstract

Multiple-Point statistics are a class of geostatistical simulation techniques used to reproduce complex spatial structures present in a training image and can be used for conditional and unconditional simulations. A weakness in multiple-point statistics is the presence of verbatim copy in realizations. Verbatim copy occurs when a continuous part of the input is ’copied’ to the simulation. Verbatim copy can result in a large part of the simulation containing identical patches of the input image, which is undesirable. This thesis has focused on the research of developing new metrics that can summarize the presence of verbatim copy in pixel-based simulations. New methods are presented which can summarize the proportion of verbatim copy. Artificial verbatim copy can be detected with these techniques, and promising results are shown on real simulation data. Both long- and short-range verbatim copy can be detected using the new methods.